Artificial Intelligence in Agriculture最新文献

{"title":"Research on an orchard row centreline multipoint autonomous navigation method based on LiDAR","authors":"Chen Zhenyu ,&nbsp;Dou Hanjie ,&nbsp;Gao Yuanyuan ,&nbsp;Zhai Changyuan ,&nbsp;Wang Xiu ,&nbsp;Zou Wei","doi":"10.1016/j.aiia.2024.12.003","DOIUrl":"10.1016/j.aiia.2024.12.003","url":null,"abstract":"<div><div>Orchard intelligent equipment must perform autonomous navigation tasks along fruit tree row centrelines and headlands according to established operational requirements. The tree canopy obstructs satellite signals, limiting the accuracy and stability of the GNSS-based autonomous navigation system. This paper presents a multipoint autonomous navigation method with the orchard row centreline navigation capabilities by integrating light detection and ranging (LiDAR) and inertial measurement unit (IMU) data. The method begins by constructing a three-dimensional (3D) point cloud map of the orchard via the LIO_SAM algorithm, and a 3D point cloud-to-two-dimensional (2D) grid map algorithm is designed. This algorithm retains the tree trunk position information from the point cloud based on tree trunk features to obtain a 2D grid map for orchard navigation, and the navigation point coordinates were calculated based on tree trunk positions. A multipoint navigation method was designed, where the system automatically determines the completion status of the previous navigation point and sequentially issues navigation point coordinates, enabling autonomous navigation along the row centrelines and headlands during orchard operations. Row centreline navigation tests and headland turning tests were conducted, and the performances of 16-line and 32-line LiDAR with this method are compared. The research results reveal that the multipoint navigation method could achieve movement along orchard row centrelines and deploy autonomous turning. The 32-line LiDAR data demonstrated an average absolute lateral deviation of 1.83 cm, a standard deviation of 1.60 cm, and a maximum deviation of 10.30 cm at a 3-m navigation point interval, indicating greater precision. However, the turning time was longer, with increases of 8.11 % and 6.13 % with the two different turning methods compared to the 16-line LiDAR. The research results provide support for research on autonomous navigation technology for intelligent orchard equipment.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"15 2","pages":"Pages 221-231"},"PeriodicalIF":8.2,"publicationDate":"2024-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143611175","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A salient feature establishment tactic for cassava disease recognition","authors":"Jiayu Zhang ,&nbsp;Baohua Zhang ,&nbsp;Zixuan Chen ,&nbsp;Innocent Nyalala ,&nbsp;Kunjie Chen ,&nbsp;Junfeng Gao","doi":"10.1016/j.aiia.2024.11.004","DOIUrl":"10.1016/j.aiia.2024.11.004","url":null,"abstract":"<div><div>Accurate classification of cassava disease, particularly in field scenarios, relies on object semantic localization to identify and precisely locate specific objects within an image based on their semantic meaning, thereby enabling targeted classification while suppressing irrelevant noise and focusing on key semantic features. The advancement of deep convolutional neural networks (CNNs) paved the way for identifying cassava diseases by leveraging salient semantic features and promising high returns. This study proposes an approach that incorporates three innovative elements to refine feature representation for cassava disease classification. First, a mutualattention method is introduced to highlight semantic features and suppress irrelevant background features in the feature maps. Second, instance batch normalization (IBN) was employed after the residual unit to construct salient semantic features using the mutualattention method, representing high-quality semantic features in the foreground. Finally, the RSigELUD activation method replaced the conventional ReLU activation, enhancing the nonlinear mapping capacity of the proposed neural network and further improving fine-grained leaf disease classification performance. This approach significantly aided in distinguishing subtle disease manifestations in cassava leaves. The proposed neural network, MAIRNet-101 (Mutualattention IBN RSigELUD Neural Network), achieved an accuracy of 95.30 % and an F1-score of 0.9531, outperforming EfficientNet-B5 and RepVGG-B3g4. To evaluate the generalization capability of MAIRNet, the FGVC-Aircraft dataset was used to train MAIRNet-50, which achieved an accuracy of 83.64 %. These results suggest that the proposed algorithm is well suited for cassava leaf disease classification applications and offers a robust solution for advancing agricultural technology.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"14 ","pages":"Pages 115-132"},"PeriodicalIF":8.2,"publicationDate":"2024-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143153819","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enhancing crop yield prediction in Senegal using advanced machine learning techniques and synthetic data","authors":"Mohammad Amin Razavi ,&nbsp;A. Pouyan Nejadhashemi ,&nbsp;Babak Majidi ,&nbsp;Hoda S. Razavi ,&nbsp;Josué Kpodo ,&nbsp;Rasu Eeswaran ,&nbsp;Ignacio Ciampitti ,&nbsp;P.V. Vara Prasad","doi":"10.1016/j.aiia.2024.11.005","DOIUrl":"10.1016/j.aiia.2024.11.005","url":null,"abstract":"<div><div>In this study, we employ advanced data-driven techniques to investigate the complex relationships between the yields of five major crops and various geographical and spatiotemporal features in Senegal. We analyze how these features influence crop yields by utilizing remotely sensed data. Our methodology incorporates clustering algorithms and correlation matrix analysis to identify significant patterns and dependencies, offering a comprehensive understanding of the factors affecting agricultural productivity in Senegal. To optimize the model's performance and identify the optimal hyperparameters, we implemented a comprehensive grid search across four distinct machine learning regressors: Random Forest, Extreme Gradient Boosting (XGBoost), Categorical Boosting (CatBoost), and Light Gradient-Boosting Machine (LightGBM). Each regressor offers unique functionalities, enhancing our exploration of potential model configurations. The top-performing models were selected based on evaluating multiple performance metrics, ensuring robust and accurate predictive capabilities. The results demonstrated that XGBoost and CatBoost perform better than the other two. We introduce synthetic crop data generated using a Variational Auto Encoder to address the challenges posed by limited agricultural datasets. By achieving high similarity scores with real-world data, our synthetic samples enhance model robustness, mitigate overfitting, and provide a viable solution for small dataset issues in agriculture. Our approach distinguishes itself by creating a flexible model applicable to various crops together. By integrating five crop datasets and generating high-quality synthetic data, we improve model performance, reduce overfitting, and enhance realism. Our findings provide crucial insights for productivity drivers in key cropping systems, enabling robust recommendations and strengthening the decision-making capabilities of policymakers and farmers in data-scarce regions.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"14 ","pages":"Pages 99-114"},"PeriodicalIF":8.2,"publicationDate":"2024-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142742953","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Neural network architecture search enabled wide-deep learning (NAS-WD) for spatially heterogenous property awared chicken woody breast classification and hardness regression","authors":"Chaitanya Pallerla ,&nbsp;Yihong Feng ,&nbsp;Casey M. Owens ,&nbsp;Ramesh Bahadur Bist ,&nbsp;Siavash Mahmoudi ,&nbsp;Pouya Sohrabipour ,&nbsp;Amirreza Davar ,&nbsp;Dongyi Wang","doi":"10.1016/j.aiia.2024.11.003","DOIUrl":"10.1016/j.aiia.2024.11.003","url":null,"abstract":"<div><div>Due to intensive genetic selection for rapid growth rates and high broiler yields in recent years, the global poultry industry has faced a challenging problem in the form of woody breast (WB) conditions. This condition has caused significant economic losses as high as $200 million annually, and the root cause of WB has yet to be identified. Human palpation is the most common method of distinguishing a WB from others. However, this method is time-consuming and subjective. Hyperspectral imaging (HSI) combined with machine learning algorithms can evaluate the WB conditions of fillets in a non-invasive, objective, and high-throughput manner. In this study, 250 raw chicken breast fillet samples (normal, mild, severe) were taken, and spatially heterogeneous hardness distribution was first considered when designing HSI processing models. The study not only classified the WB levels from HSI but also built a regression model to correlate the spectral information with sample hardness data. To achieve a satisfactory classification and regression model, a neural network architecture search (NAS) enabled a wide-deep neural network model named NAS-WD, which was developed. In NAS-WD, NAS was first used to automatically optimize the network architecture and hyperparameters. The classification results show that NAS-WD can classify the three WB levels with an overall accuracy of 95 %, outperforming the traditional machine learning model, and the regression correlation between the spectral data and hardness was 0.75, which performs significantly better than traditional regression models.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"14 ","pages":"Pages 73-85"},"PeriodicalIF":8.2,"publicationDate":"2024-11-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142699967","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Utility-based regression and meta-learning techniques for modeling actual ET: Comparison to (METRIC-EEFLUX) model","authors":"Fatima K. Abu Salem ,&nbsp;Sara Awad ,&nbsp;Yasmine Hamdar ,&nbsp;Samer Kharroubi ,&nbsp;Hadi Jaafar","doi":"10.1016/j.aiia.2024.11.001","DOIUrl":"10.1016/j.aiia.2024.11.001","url":null,"abstract":"<div><div>Estimating actual evapotranspiration (ETₐ) is crucial for water resource management, yet existing methods face limitations. Traditional approaches, including eddy covariance and remote sensing-based energy balance methods, often struggle with high costs, limited spatial and temporal coverage, and reduced predictive accuracy, particularly for classical empirical models. While machine learning has emerged as a promising alternative, it still presents challenges, notably in underestimating ETₐ during periods of high heat. We attribute this to insufficient learning on the rare but highly relevant ETₐ values of interest, or the not-so-big climatic datasets available for use. In this manuscript, we demonstrate how <em>few-shot, meta-learning models (MAML)</em> that are specifically designed for enhanced generalizability on not-so-big datasets can outperform basic machine learning models in upscaling ETₐ from two major in-situ towers, the Ameriflux and Euroflux. Using limited remotely sensed land surface data from the METRIC-EEFlux and limited climatic variables, we demonstrate that the chosen models can attain quantifiable utility within the <em>utility-based-regression</em> paradigm towards impactful practical considerations. Our initial explorations reveal that EEflux ETₐ deviates significantly from in-situ observations measured through the Ameriflux and EEflux towers (<span><math><msup><mi>R</mi><mn>2</mn></msup><mo>=</mo><mn>39</mn><mo>%</mo></math></span>). Instead, MAML shows best performance in approximating ETₐ than basic machine learning algorithms and EEFlux (<span><math><msup><mi>R</mi><mn>2</mn></msup><mo>=</mo><mn>71</mn><mo>%</mo></math></span> on entire testing dataset, <span><math><msup><mi>R</mi><mn>2</mn></msup><mo>=</mo><mn>0.88</mn></math></span> on the Csa climate, <span><math><msup><mi>R</mi><mn>2</mn></msup><mo>=</mo><mn>0.79</mn></math></span> on the Cfa climate, and <span><math><msup><mi>R</mi><mn>2</mn></msup><mo>=</mo><mn>0.78</mn></math></span> on the CSH vegetation class), and continues to improve without overfitting even when exposed to a relatively small training dataset. Its high F2 score (96 %) indicates that MAML has very high precision and recall for rare cases, which is significant for irrigation. Of independent interest, this study confirms that limited remotely sensed EEflux products contribute significantly to knowledge about ground truth ETₐ and can thus be of valuable use in settings where access to good quality and high-volume data is compromised.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"14 ","pages":"Pages 43-55"},"PeriodicalIF":8.2,"publicationDate":"2024-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142699966","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Detectability of multi-dimensional movement and behaviour in cattle using sensor data and machine learning algorithms: Study on a Charolais bull","authors":"Miklós Biszkup ,&nbsp;Gábor Vásárhelyi ,&nbsp;Nuri Nurlaila Setiawan ,&nbsp;Aliz Márton ,&nbsp;Szilárd Szentes ,&nbsp;Petra Balogh ,&nbsp;Barbara Babay-Török ,&nbsp;Gábor Pajor ,&nbsp;Dóra Drexler","doi":"10.1016/j.aiia.2024.11.002","DOIUrl":"10.1016/j.aiia.2024.11.002","url":null,"abstract":"<div><div>The development of motion sensors for monitoring cattle behaviour has enabled farmers to predict the state of their cattle's welfare more efficiently. While most studies work with one dimensional output with disjunct behaviour categories, more accurate prediction can still be achieved by including complex movements and enriching the sensor algorithm to detect multi-dimensional movements, i.e., more than one movement occurring simultaneously. This paper presents such a machine-learning method for analysing overlapping independent movements. The output of the method consists of automatically recognized complex behaviour patterns that can be used for measuring animal welfare, predicting calving, or detecting early signs of diseases. This study combines automated motion sensors (i.e., halter and pedometer) for ruminants known as RumiWatch mounted on a Charolais fattening bull and camera observation. Fourteen types of complex movements were identified, i.e., defecating-urinating, eating, drinking, getting up, head movement, licking, lying down, lying, playing-aggression, rubbing, ruminating, sleeping, standing, and stepping. As multiple parallel binary classificators were used, the system was able to recognize parallel behavioural patterns with high fidelity. Two types of machine learning, i.e., Support Vector Classification (SVC) and RandomForest were used to recognize different general and non-general forms of movement. Results from these two supervised learning systems were compared. A continuous forty-eight hours of video were annotated to train the systems and validate their predictions. The success rate of both classifiers in recognizing special movements from both sensors or separately in different settings (i.e., window and padding) was examined. Although the two classifiers produced different results, the ideal settings showed that all forms of movement in the subject animal were successfully recognized with high accuracy. More studies using more individual animals and different ruminants would increase our knowledge on enhancing the system's performance and accuracy.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"14 ","pages":"Pages 86-98"},"PeriodicalIF":8.2,"publicationDate":"2024-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142699968","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Estimating TYLCV resistance level using RGBD sensors in production greenhouse conditions","authors":"Dorin Shmaryahu ,&nbsp;Rotem Lev Lehman ,&nbsp;Ezri Peleg ,&nbsp;Guy Shani","doi":"10.1016/j.aiia.2024.10.004","DOIUrl":"10.1016/j.aiia.2024.10.004","url":null,"abstract":"<div><div>Automated phenotyping is the task of automatically measuring plant attributes to help farmers and breeders in developing and growing strong robust plants. An automated tool for early illness detection can accelerate the process of identifying plant resistance and quickly pinpoint problematic breeding. Many such phenotyping tasks can be achieved by analyzing images from simple, low cost, RGB-D sensors. In this paper we focused on a particular case study — identifying the resistance level of tomato hybrids to the tomato yellow leaf curl virus (TYLCV) in production greenhouses. This is a difficult task, as separating between resistance levels based on images is difficult even for expert breeders. We collected a large dataset of images from an experiment containing many tomato hybrids with varying resistance levels. We used the depth information to identify the topmost part of the tomato plant. We then used deep learning models to classify the various resistance levels. For identifying plants with visual symptoms, our methods achieved an accuracy of 0.928, a precision of 0.934, and a recall of 0.95. In the multi-class case we achieved an accuracy of 0.76 in identifying the correct level, and an error of 0.278. Our methods are not particularly tailored for the specific task, and can be extended to other tasks that identify various plant diseases with visual symptoms such as ToBRFV, mildew, ToMV and others.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"14 ","pages":"Pages 31-42"},"PeriodicalIF":8.2,"publicationDate":"2024-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142699964","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Development of a cutting-edge ensemble pipeline for rapid and accurate diagnosis of plant leaf diseases","authors":"S.M. Nuruzzaman Nobel ,&nbsp;Maharin Afroj ,&nbsp;Md Mohsin Kabir ,&nbsp;M.F. Mridha","doi":"10.1016/j.aiia.2024.10.005","DOIUrl":"10.1016/j.aiia.2024.10.005","url":null,"abstract":"<div><div>Selecting techniques is a crucial aspect of disease detection analysis, particularly in the convergence of computer vision and agricultural technology. Maintaining crop disease detection in a timely and accurate manner is essential to maintaining global food security. Deep learning is a viable answer to meet this need. To proceed with this study, we have developed and evaluated a disease detection model using a novel ensemble technique. We propose to introduce DenseNetMini, a smaller version of DenseNet. We propose combining DenseNetMini with a learning resizer in ensemble approach to enhance training accuracy and expedite learning. Another unique proposition involves utilizing Gradient Product (GP) as an optimization technique, effectively reducing the training time and improving the model performance. Examining images at different magnifications reveals noteworthy diagnostic agreement and accuracy improvements. Test accuracy rates of 99.65 %, 98.96 %, and 98.11 % are seen in the Plantvillage, Tomato leaf, and Appleleaf9 datasets, respectively. One of the research's main achievements is the significant decrease in processing time, which suggests that using the GP could improve disease detection in agriculture's accessibility and efficiency. Beyond quantitative successes, the study highlights Explainable Artificial Intelligence (XAI) methods, which are essential to improving the disease detection model's interpretability and transparency. XAI enhances the interpretability of the model by visually identifying critical areas on plant leaves for disease identification, which promotes confidence and understanding of the model's functionality.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"14 ","pages":"Pages 56-72"},"PeriodicalIF":8.2,"publicationDate":"2024-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142699965","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A review of external quality inspection for fruit grading using CNN models","authors":"Luis E. Chuquimarca ,&nbsp;Boris X. Vintimilla ,&nbsp;Sergio A. Velastin","doi":"10.1016/j.aiia.2024.10.002","DOIUrl":"10.1016/j.aiia.2024.10.002","url":null,"abstract":"<div><div>This article reviews the state of the art of recent CNN models used for external quality inspection of fruits, considering parameters such as color, shape, size, and defects, used to categorize fruits according to international marketing levels of agricultural products. The literature review considers the number of fruit images in different datasets, the type of images used by the CNN models, the performance results obtained by each CNNs, the optimizers that help increase the accuracy of these, and the use of pre-trained CNN models used for transfer learning. CNN models have used various types of images in the visible, infrared, hyperspectral, and multispectral bands. Furthermore, the fruit image datasets used are either real or synthetic. Finally, several tables summarize the articles reviewed, which are prioritized according to inspection parameters, facilitating a critical comparison of each work.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"14 ","pages":"Pages 1-20"},"PeriodicalIF":8.2,"publicationDate":"2024-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142533851","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Classifying early apple scab infections in multispectral imagery using convolutional neural networks","authors":"Alexander J. Bleasdale,&nbsp;J. Duncan Whyatt","doi":"10.1016/j.aiia.2024.10.001","DOIUrl":"10.1016/j.aiia.2024.10.001","url":null,"abstract":"<div><div>Multispectral imaging systems combined with deep learning classification models can be cost-effective tools for the early detection of apple scab (<em>Venturia inaequalis</em>) disease in commercial orchards. Near-infrared (NIR) imagery can display apple scab symptoms earlier and at a greater severity than visible-spectrum (RGB) imagery. Early apple scab diagnosis based on NIR imagery may be automated using deep learning convolutional neural networks (CNNs). CNN models have previously been used to classify a range of apple diseases accurately but have primarily focused on identifying late-stage rather than early-stage detection. This study fine-tunes CNN models to classify apple scab symptoms as they progress from the early to late stages of infection using a novel multispectral (RGB-NIR) time series created especially for this purpose.</div><div>This novel multispectral dataset was used in conjunction with a large Apple Disease Identification (ADID) dataset created from publicly available, pre-existing disease datasets. This ADID dataset contained 29,000 images of infection symptoms across six disease classes. Two CNN models, the lightweight MobileNetV2 and heavyweight EfficientNetV2L, were fine-tuned and used to classify each disease class in a testing dataset, with performance assessed through metrics derived from confusion matrices. The models achieved scab-prediction accuracies of 97.13 % and 97.57 % for MobileNetV2 and EfficientNetV2L, respectively, on the secondary data but only achieved accuracies of 74.12 % and 78.91 % when applied to the multispectral dataset in isolation. These lower performance scores were attributed to a higher proportion of false-positive scab predictions in the multispectral dataset. Time series analyses revealed that both models could classify apple scab infections earlier than the manual classification techniques, leading to more false-positive assessments, and could accurately distinguish between healthy and infected samples up to 7 days post-inoculation in NIR imagery.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"15 1","pages":"Pages 39-51"},"PeriodicalIF":8.2,"publicationDate":"2024-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143097984","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}